Molecular mechanism of Gαi activation by non-GPCR proteins with a Gα-Binding and Activating motif

Heterotrimeric G proteins are quintessential signalling switches activated by nucleotide exchange on Gα. Although activation is predominantly carried out by G-protein-coupled receptors (GPCRs), non-receptor guanine-nucleotide exchange factors (GEFs) have emerged as critical signalling molecules and therapeutic targets. Here we characterize the molecular mechanism of G-protein activation by a family of non-receptor GEFs containing a Gα-binding and -activating (GBA) motif. We combine NMR spectroscopy, computational modelling and biochemistry to map changes in Gα caused by binding of GBA proteins with residue-level resolution. We find that the GBA motif binds to the SwitchII/α3 cleft of Gα and induces changes in the G-1/P-loop and G-2 boxes (involved in phosphate binding), but not in the G-4/G-5 boxes (guanine binding). Our findings reveal that G-protein-binding and activation mechanisms are fundamentally different between GBA proteins and GPCRs, and that GEF-mediated perturbation of nucleotide phosphate binding is sufficient for Gα activation.

spectra depicting the perturbations in the same Gαi3 signals induced by GIV binding as shown in Figure 2A. B. Quantification of GIV-induced NMR perturbations on Gαi3. Chemical shift (CSP, top graph) or intensity perturbations (I ratio , bottom graph) of the backbone amide signals of the TROSY NMR spectra in panel A. Red, orange and yellow circles indicate residues undergoing perturbations larger than 10, 5 or 3 times the median (M), respectively. Blue circles indicate Gαi3 residues with perturbations smaller than 3 times the median (M) and grey circles residues for which no reliable NMR measurement could be made. The horizontal black bar in the middle depicts the secondary structure elements of Gαi3 and is annotated with the position of the 3 switch regions (green) that undergo dramatic conformational changes upon GTP binding and the 5 conserved G-box sequences (blue) that mediate nucleotide binding. generated by the binding of the non-hydrolyzable GTP analog GTPγS adopts a conformation that is resistant to trypsin digestion outside of a short N-terminal sequence which is cleaved off by trypsin 9 . All Gαi3 mutants adopt a trypsin-resistant conformation upon GTPγS binding (indicated by * on the right) to a similar extent as that observed for Gαi3 WT, indicating that the mutants do not have overt folding defects that preclude GTP binding and activation. Only Gαi3 L249H showed a modest decrease in trypsin protection but its steady-state GTPase activity was intact (see Fig. Supplementary 6B). One experiment representative of at least 3 is shown. C. Full dataset with loading and negative controls for the pulldown results shown in Fig.  4 and Fig. 6. Binding of His-Gαi3 WT or mutants to GST-GIV (aa1671-1755) (top), GST-DAPLE (residues 1650-1880) (middle) or GST (bottom) immobilized on glutathione-agarose beads was determined in pulldown assays. Resin-bound proteins were eluted, separated by SDS-PAGE and analyzed by Coomassie blue staining. The multiple band patterns of GST-GIV and GST-DAPLE correspond to degradation products generated during protein purification. High sensitivity to proteolytic degradation is consistent with the disordered nature of these fragments (Supplementary Fig. 4 and see reference 14 ) but it is not a concern for the conclusions drawn about the binding of different Gαi3 mutants because (i) binding of Gαi3 WT to the same GST-GIV preparation (i.e., with identical proteolytic cleavage pattern) was used as an internal control for each gel/ experiment and, (ii) equivalent G protein binding is observed with preparations displaying less degradation products (Fig. 1). No binding to GST was detected. One experiment representative of at least 3 is shown. The green and red dashed boxes indicate the datasets corresponding to the results shown in Fig. 4 and Fig. 6, respectively. The result with His-Gαi3 WT is reproduced in every graph (black traces) to facilitate comparison with the mutants. Mean ± S.E.M, n=3-7.

B. GTPase and GTPγS binding activities for Gαi3 mutants used in this study. Basal
GTPase and GTPγS binding activities of His-Gαi3 WT and the indicated mutants (first column) in the absence of GIV or DAPLE were determined as described in Methods. The second and fourth columns correspond to the raw GTPase and GTPγS binding activities, respectively (mean ± S.E.M, n≥3). The third and fifth columns correspond to GTPase and GTPγS binding activities expressed relative to WT (%). ND= Not determined. Mutants with GTPase activities below 65% of that of WT (marked in red in column two) were excluded from subsequent analyses of Gαi3 activation by GIV or DAPLE (they correspond to the ND, not determined mutants in the GTPase assays shown in Fig. 4 and Fig. 6). GTPase defects of selected mutants correlated well with defects in the basal GTPγS binding (column five), indicating impaired nucleotide binding rather impaired nucleotide hydrolysis. C. Effect of GIV and DAPLE on the rate of GTPγS binding by selected Gαi3 mutants. GTPγS binding to His-Gαi3 WT or the indicated mutants in the absence (white) or presence of GIV (black) or DAPLE (grey) was determined as described in Methods. Mean ± S.E.M, n=3. The lack of GIV-or DAPLE-mediated activation for mutants W211A and N256E in GTPγS binding assays correlates with the lack of activation in steady-state GTPase assays shown in Fig. 4. Conversely, GIV and DAPLE enhance GTPγS binding to the Gαi3 ΔC9 mutant as efficiently as for WT, which correlates with the activation observed for the same mutant in steady-state GTPase assays shown in Fig. 6.   Supplementary Fig. 8. Characterization of the structural determinants of GIV required for Gαi3 binding. A. View of GIV docked on Gαi3 and localization of GIV residues selected for systematic peptide array mutagenesis. Left, Representation of the molecular model of Gαi3 bound to GIVpept (green ribbon). The surface of Gαi3 is colored based on NMR signal perturbations, and was generated as described in Fig. 3. Right, Gαi3 was removed from the image on the left and the residues (sticks) subjected to peptide array-based mutagenesis labeled. B. Mutagenesis scanning of selected GIV residues in peptide arrays. 24-mer GIV peptides corresponding to the sequence shown on top were synthesized and immobilized on slides. Each one of the residues indicated in red in the sequence on top was substituted with every other natural amino acid. Each row corresponds to a series of mutants in which the residue indicated on the left is mutated to the amino acid type indicated on the top. Each spot is one peptide and the spots circled in white correspond to the wild-type peptide in each series. The immobilized peptides were probed in batch with purified Gαi3 and binding determined after sequential incubation with primary and secondary antibodies coupled to fluorescent probes. Similar results were obtained in two other experiments, and similar peptide content in all the spots was validated by Coomassie staining. See Supplementary Note 2 for further discussion of these results in the context of the Gαi3 mutagenesis data. C. Q1683S or E1677S mutation in GIV impairs Gαi3 binding in pulldown assays. Q1683S and E1677S were selected for validation in protein-protein binding experiments in solution (see Supplementary Note 2 for the rationale of residue selection). Binding of His-Gαi3 WT or mutants to GST-GIV (residues 1671-1755) or GST was determined in pulldown assays. Resinbound proteins were eluted, separated by SDS-PAGE and analyzed by Coomassie blue staining. The multiple band pattern of GST-GIV corresponds to degradation products generated during protein purification. No binding to GST was detected. One experiment of at least 3 is shown. Figure 9. Images of uncropped gels. Red boxes indicate the parts of that are shown in the corresponding final figures.

SUPPLEMENTARY NOTE 1
Rationale for the design of Gαi3 mutants-The mutants used in Fig. 4 and Fig.6 were designed based on the joint analysis of the NMR signal perturbations upon GIV binding (Fig. 4,   Supplementary Fig. 3) and the in silico prediction of the contribution of individual amino acids to the energetics of the GIV-Gαi3 interaction (Fig. 4A, 4B, Fig. 6A). First, we looked for amino acids displaying NMR signal perturbations that were representative of different regions of Gαi3 (β1 strand, P-loop, α3 helix, α3/α5 loop, α4/β6 loop and SwII and SwI regions). Within this group, we selected amino acids that were predicted to stabilize the interaction with GIV (K35, W211, F215, L249, N256, W258, F259) or not (L36, L37, G42, I184, V218, S252, R313, K317).
We also mutated L39 and K257, two residues for which no NMR information was available but were adjacent to other amino acids undergoing NMR signal perturbations. These were chosen to test the accuracy of our homology model and related in silico predictions. L39 was mutated because our model predicted it to stabilize GIV binding despite being adjacent to two amino acids (L37 and G40) not predicted to contribute to GIV binding (but displaying strong NMR signal perturbations). K257 was mutated because our model predicted it to not stabilize GIV binding despite being adjacent to two amino acids (N256 and W258) predicted to contribute to GIV binding and/or displaying strong NMR signal perturbations. The default mutation was to alanine (A) but in some cases residues were mutated to other amino acids based on a closer examination of the homology model because they were expected to have a more marked effect on binding. These include L249V, L249H, S252D, N256E and K317W. W258 was mutated to F to serve as an internal benchmark because the effect of this mutation on GIV binding has been extensively characterized by different techniques. W211A and F215A were also included as benchmarks because both of them have been shown to impair GIV 10 , DAPLE 15 or Calnuc 16 binding using experimental approaches different from those described here. Supplementary Fig. 8-To further validate the results derived from the Gαi3 mutagenesis and support that our homology model faithfully represents the structural features of the GIV-Gαi3 interaction, we performed a systematic mutagenesis analysis of the GBA motif of GIV. We reasoned that those amino acid positions more directly involved in binding Gαi3 would be more sensitive to substitution by other amino acids and/or by the chemical nature of the side chain substitution. Nine positions in the region of the GBA motif were chosen for this analysis. These included positions predicted to participate in (e.g., Q1683) or not (e.g., V1679), as well as positions already known to be important for binding as benchmarks (e.g., F1685 10,17 ). Each position was mutated to every other natural amino acid to create a grid of 171 immobilized peptide variants that were probed in batch for Gαi3 binding. The results obtained through this approach complemented the results of the Gαi3 mutagenesis studies (Fig. 4) and validated the accuracy of our model. For example, F1685 has been repeatedly shown to be crucial for Gαi3 binding 10,17 and the peptide array revealed that its substitution by any other amino acid dramatically impairs G protein binding.

Rationale and interpretation of the experiments shown in
Similarly, L1682 and L1686, two residues that together with F1685 dock onto the hydrophobic cleft lined by Gαi3 residues W211 and F215 based on our model, are also very sensitive to mutation. The GIV-Gαi3 interaction only tolerates some substitutions for other hydrophobic residues in these positions, which is also consistent with our model. Mutation of E1688 to A has been previously shown to disrupt GIV binding to Gαi3 10  showing that V1679 is predominantly solvent exposed, the peptide array revealed that this position can be replaced by virtually any other amino acid without impairing Gαi3 binding. In contrast, V1680 can only be replaced by other hydrophobic residues without affecting Gαi3 binding, which is consistent with our model showing that this residue is buried in a hydrophobic environment of Gαi3. Many mutations in positions Q1683 and E1687 also impair Gαi3 binding, which is consistent with our results with Gαi3 mutants (Fig. 4). For example, Q1683 is located in the vicinity of Gαi3 L249 and E1687 in the vicinity of Gαi3 S252/N256. Because the role of these two positions of GIV on Gαi3 binding has not been previously investigated, we performed additional validation experiments to evaluate protein-protein binding in solution. For this, we mutated each one of these residues to serine (Q1683S or E1687S). We chose serine because it is one of the substitutions that impair Gαi3 binding in the peptide array format yet it is moderately conservative. We found that both mutants impaired GIV binding to Gαi3 as determined by GST pulldown assays (Supplementary Fig. 8C).